=================================================================== RCS file: /cvs/pod2mdoc/pod2mdoc.c,v retrieving revision 1.8 retrieving revision 1.13 diff -u -p -r1.8 -r1.13 --- pod2mdoc/pod2mdoc.c 2014/03/23 23:57:38 1.8 +++ pod2mdoc/pod2mdoc.c 2014/04/01 13:05:50 1.13 @@ -1,4 +1,4 @@ -/* $Id: pod2mdoc.c,v 1.8 2014/03/23 23:57:38 kristaps Exp $ */ +/* $Id: pod2mdoc.c,v 1.13 2014/04/01 13:05:50 kristaps Exp $ */ /* * Copyright (c) 2014 Kristaps Dzonsons * @@ -26,6 +26,11 @@ #include #include +/* + * In what section can we find Perl manuals? + */ +#define PERL_SECTION "3p" + struct args { const char *title; /* override "Dt" title */ const char *date; /* override "Dd" date */ @@ -39,11 +44,17 @@ enum list { LIST__MAX }; +enum sect { + SECT_NONE = 0, + SECT_NAME, /* NAME section */ + SECT_SYNOPSIS, /* SYNOPSIS section */ +}; + struct state { int parsing; /* after =cut of before command */ int paused; /* in =begin and before =end */ int haspar; /* in paragraph: do we need Pp? */ - int isname; /* are we the NAME section? */ + enum sect sect; /* which section are we in? */ const char *fname; /* file being parsed */ #define LIST_STACKSZ 128 enum list lstack[LIST_STACKSZ]; /* open lists */ @@ -158,6 +169,131 @@ formatescape(const char *buf, size_t *start, size_t en } /* + * Run some heuristics to intuit a link format. + * I recognise L as a Perl manpage, printing it in section 3p; + * or a general UNIX foo(5) manpage. + * If I recognise one, I set "start" to be the end of the sequence so + * that the caller can safely just continue processing. + * Otherwise, I don't touch "start". + */ +static int +trylink(const char *buf, size_t *start, size_t end, size_t dsz) +{ + size_t sv, nstart, nend, i, j; + int hasdouble; + + /* + * Scan to the start of the terminus. + * This function is more or less replicated in the formatcode() + * for null or index formatting codes. + */ + hasdouble = 0; + for (sv = nstart = *start; nstart < end; nstart++) { + /* Do we have a double-colon? */ + if (':' == buf[nstart] && + nstart > sv && + ':' == buf[nstart - 1]) + hasdouble = 1; + if ('>' != buf[nstart]) + continue; + else if (dsz == 1) + break; + assert(nstart > 0); + if (' ' != buf[nstart - 1]) + continue; + i = nstart; + for (j = 0; i < end && j < dsz; j++) + if ('>' != buf[i++]) + break; + if (dsz == j) + break; + } + + /* We don't care about stubs. */ + if (nstart == end || nstart == *start) + return(0); + + /* Set nend to the end of content. */ + nend = nstart; + if (dsz > 1) + nend--; + + /* + * Provide for some common invocations of the link primitive. + * First, allow us to link to other Perl manuals. + */ + if (hasdouble) + printf("Xr %.*s " PERL_SECTION, + (int)(nend - sv), &buf[sv]); + else if (nend - sv > 3 && isalnum(buf[sv]) && + ')' == buf[nend - 1] && + isdigit((int)buf[nend - 2]) && + '(' == buf[nend - 3]) + printf("Xr %.*s %c", + (int)(nend - 3 - sv), + &buf[sv], buf[nend - 2]); + else + return(0); + + *start = nstart; + return(1); +} + + +/* + * Doclifting: if we're a bold "-xx" and we're in the SYNOPSIS section, + * then it's likely that we're a flag. + * Our flag might be followed by an argument, so make sure that we're + * accounting for that, too. + * If we don't have a flag at all, however, then assume we're an "Ar". + */ +static void +dosynopsisfl(const char *buf, size_t *start, size_t end) +{ + size_t i; +again: + (*start)++; + for (i = *start; i < end; i++) + if (isalnum((int)buf[i])) + continue; + else if ('-' == buf[i]) + continue; + else if ('_' == buf[i]) + continue; + else + break; + + assert(i < end); + + if ( ! (' ' == buf[i] || '>' == buf[i])) { + printf("Ar "); + return; + } + + printf("Fl "); + if (end - *start > 1 && + isupper((int)buf[*start]) && + islower((int)buf[*start + 1]) && + (end - *start == 2 || + ' ' == buf[*start + 2])) + printf("\\&"); + printf("%.*s ", (int)(i - *start), &buf[*start]); + *start = i; + + if (' ' == buf[i]) { + while (i < end && ' ' == buf[i]) + i++; + assert(i < end); + if ('-' == buf[i]) { + *start = i; + goto again; + } + printf("Ar "); + *start = i; + } +} + +/* * We're at the character in front of a format code, which is structured * like X<...> and can contain nested format codes. * This consumes the whole format code, and any nested format codes, til @@ -171,8 +307,8 @@ formatescape(const char *buf, size_t *start, size_t en * reentrant). */ static int -formatcode(const char *buf, size_t *start, - size_t end, int reentrant, int nomacro) +formatcode(struct state *st, const char *buf, + size_t *start, size_t end, int reentrant, int nomacro) { enum fmt fmt; size_t i, j, dsz; @@ -278,19 +414,24 @@ formatcode(const char *buf, size_t *start, printf("Em "); break; case (FMT_BOLD): - printf("Sy "); + if (SECT_SYNOPSIS == st->sect && + 1 == dsz && + '-' == buf[*start]) + dosynopsisfl(buf, start, end); + else + printf("Sy "); break; case (FMT_CODE): printf("Qo Li "); break; case (FMT_LINK): - printf("Lk "); + if ( ! trylink(buf, start, end, dsz)) + printf("No "); break; case (FMT_FILE): printf("Pa "); break; case (FMT_NBSP): - /* TODO. */ printf("No "); break; default: @@ -326,7 +467,7 @@ formatcode(const char *buf, size_t *start, } } if (*start + 1 < end && '<' == buf[*start + 1]) { - formatcode(buf, start, end, 1, nomacro); + formatcode(st, buf, start, end, 1, nomacro); continue; } @@ -381,13 +522,14 @@ formatcode(const char *buf, size_t *start, * Calls formatcode() til the end of a paragraph. */ static void -formatcodeln(const char *buf, size_t *start, size_t end, int nomacro) +formatcodeln(struct state *st, const char *buf, + size_t *start, size_t end, int nomacro) { last = ' '; while (*start < end) { if (*start + 1 < end && '<' == buf[*start + 1]) { - formatcode(buf, start, end, 1, nomacro); + formatcode(st, buf, start, end, 1, nomacro); continue; } /* @@ -487,24 +629,28 @@ command(struct state *st, const char *buf, size_t star * how pod2man handles it. */ printf(".Sh "); - st->isname = 0; - if (end - start == 4) + st->sect = SECT_NONE; + if (end - start == 4) { if (0 == memcmp(&buf[start], "NAME", 4)) - st->isname = 1; - formatcodeln(buf, &start, end, 1); + st->sect = SECT_NAME; + } else if (end - start == 8) { + if (0 == memcmp(&buf[start], "SYNOPSIS", 8)) + st->sect = SECT_SYNOPSIS; + } + formatcodeln(st, buf, &start, end, 1); putchar('\n'); st->haspar = 1; break; case (CMD_HEAD2): printf(".Ss "); - formatcodeln(buf, &start, end, 1); + formatcodeln(st, buf, &start, end, 1); putchar('\n'); st->haspar = 1; break; case (CMD_HEAD3): puts(".Pp"); printf(".Em "); - formatcodeln(buf, &start, end, 0); + formatcodeln(st, buf, &start, end, 0); putchar('\n'); puts(".Pp"); st->haspar = 1; @@ -512,7 +658,7 @@ command(struct state *st, const char *buf, size_t star case (CMD_HEAD4): puts(".Pp"); printf(".No "); - formatcodeln(buf, &start, end, 0); + formatcodeln(st, buf, &start, end, 0); putchar('\n'); puts(".Pp"); st->haspar = 1; @@ -566,7 +712,7 @@ command(struct state *st, const char *buf, size_t star switch (st->lstack[st->lpos - 1]) { case (LIST_TAG): printf(".It "); - formatcodeln(buf, &start, end, 0); + formatcodeln(st, buf, &start, end, 0); putchar('\n'); break; case (LIST_ENUM): @@ -646,6 +792,61 @@ verbatim(struct state *st, const char *buf, size_t sta } /* + * See dosynopsisop(). + */ +static int +hasmatch(const char *buf, size_t start, size_t end) +{ + size_t stack; + + for (stack = 0; start < end; start++) + if (buf[start] == '[') + stack++; + else if (buf[start] == ']' && 0 == stack) + return(1); + else if (buf[start] == ']') + stack--; + return(0); +} + +/* + * If we're in the SYNOPSIS section and we've encounter braces in an + * ordinary paragraph, then try to see whether we're an [-option]. + * Do this, if we're an opening bracket, by first seeing if we have a + * matching end via hasmatch(). + * If we're an ending bracket, see if we have a stack already. + */ +static int +dosynopsisop(const char *buf, int *last, + size_t *start, size_t end, size_t *opstack) +{ + + assert('[' == buf[*start] || ']' == buf[*start]); + + if ('[' == buf[*start] && hasmatch(buf, *start + 1, end)) { + if ('\n' != *last) + putchar('\n'); + puts(".Oo"); + (*opstack)++; + } else if ('[' == buf[*start]) + return(0); + + if (']' == buf[*start] && *opstack > 0) { + if ('\n' != *last) + putchar('\n'); + puts(".Oc"); + (*opstack)--; + } else if (']' == buf[*start]) + return(0); + + (*start)++; + *last = '\n'; + while (' ' == buf[*start]) + (*start)++; + return(1); +} + +/* * Ordinary paragraph. * Well, this is really the hardest--POD seems to assume that, for * example, a leading space implies a newline, and so on. @@ -656,7 +857,7 @@ verbatim(struct state *st, const char *buf, size_t sta static void ordinary(struct state *st, const char *buf, size_t start, size_t end) { - size_t i, j; + size_t i, j, opstack; if ( ! st->parsing || st->paused) return; @@ -667,7 +868,7 @@ ordinary(struct state *st, const char *buf, size_t sta * we're in "name - description" format. * To wit, print out a "Nm" and "Nd" in that format. */ - if (st->isname) { + if (SECT_NAME == st->sect) { for (i = end - 1; i > start; i--) if ('-' == buf[i]) break; @@ -678,11 +879,11 @@ ordinary(struct state *st, const char *buf, size_t sta if ('-' != buf[i]) break; printf(".Nm "); - formatcodeln(buf, &start, i + 1, 1); + formatcodeln(st, buf, &start, i + 1, 1); putchar('\n'); start = j + 1; printf(".Nd "); - formatcodeln(buf, &start, end, 1); + formatcodeln(st, buf, &start, end, 1); putchar('\n'); return; } @@ -693,6 +894,7 @@ ordinary(struct state *st, const char *buf, size_t sta st->haspar = 0; last = '\n'; + opstack = 0; while (start < end) { /* @@ -708,6 +910,17 @@ ordinary(struct state *st, const char *buf, size_t sta printf("\\&"); else if ('\n' == last && '\'' == buf[start]) printf("\\&"); + /* + * If we're in the SYNOPSIS, have square + * brackets indicate that we're opening and + * closing an optional context. + */ + if (SECT_SYNOPSIS == st->sect && + ('[' == buf[start] || + ']' == buf[start]) && + dosynopsisop(buf, &last, + &start, end, &opstack)) + continue; putchar(last = buf[start++]); if ('\\' == last) putchar('e'); @@ -723,7 +936,7 @@ ordinary(struct state *st, const char *buf, size_t sta * Consume all whitespace so we don't * accidentally start an implicit literal line. */ - if (formatcode(buf, &start, end, 0, 0)) { + if (formatcode(st, buf, &start, end, 0, 0)) { putchar(last = '\n'); while (start < end && ' ' == buf[start]) start++; @@ -752,19 +965,7 @@ ordinary(struct state *st, const char *buf, size_t sta for ( ; start < end; start++) if (' ' != buf[start] && '\t' != buf[start]) break; - } else if (start < end) { - /* - * Default: print the character. - * Escape initial control characters. - */ - if ('\n' == last && '.' == buf[start]) - printf("\\&"); - else if ('\n' == last && '\'' == buf[start]) - printf("\\&"); - putchar(last = buf[start++]); - if ('\\' == last) - putchar('e'); - } + } } if (last != '\n') @@ -828,7 +1029,7 @@ dofile(const struct args *args, const char *fname, if (NULL != (cp = strrchr(title, '.'))) { *cp++ = '\0'; if (0 == strcmp(cp, "pm")) - section = "3p"; + section = PERL_SECTION; } }